[glib/th/g-ptr-array-set-null-terminated] fixup! array: add support for g_ptr_array_null_terminated()

[Thomas Haller <thaller src gnome org>]

commit c7252b356c65e635972a002eb2dfaa526bfae4a3
Author: Thomas Haller <thaller redhat com>
Date:   Fri May 15 16:58:18 2020 +0200

    fixup! array: add support for g_ptr_array_null_terminated()
    
    https://gitlab.gnome.org/GNOME/glib/-/merge_requests/1485#note_800510
    
    > I think we should guarantee that g_ptr_array_free() always returns a
    > non-NULL array if it’s supposed to be NULL-terminated, but we shouldn’t
    > worry about the pdata if the GPtrArray is still reffed. That can be
    > cleared to NULL, as is already the case after constructing a new array.
    
    th: I disagree. The documentation already makes it clear that an empty array
      may either be represented as NULL or [ NULL ]. It doesn't seem useful
      to artificially allocate an empty array during g_ptr_array_free(). If
      the user cares, she can just do
    
          return g_ptr_array_free (parray, FALSE) ?: g_new0 (char *, 1);
    
      Anyway, here is a fixup implementing that.

 glib/garray.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
---
diff --git a/glib/garray.c b/glib/garray.c
index 9cc53f17d..4f90407d1 100644
--- a/glib/garray.c
+++ b/glib/garray.c
@@ -1549,7 +1549,12 @@ ptr_array_free (GPtrArray      *array,
       segment = NULL;
     }
   else
-    segment = rarray->pdata;
+    {
+      segment = rarray->pdata;
+      if (   !segment
+          && rarray->null_terminated)
+        segment = g_new0 (gpointer, 1);
+    }
 
   if (flags & PRESERVE_WRAPPER)
     {